Land degradation degree classification method based on convolutional neural network and multi-source data

Abstract

The invention provides a land degradation degree classification method based on a convolutional neural network and multi-source data. The method comprises the following steps: S1, determining a land degradation degree standard; s2, acquiring land degradation degree sample point data and performing sample point classification according to a land degradation degree standard; s3, acquiring environment variable data and preprocessing the environment variable data; s4, performing unified processing on the classified land degradation degree sample point data and the preprocessed environment variable data to obtain sample data; s5, constructing a land degradation degree classification model and training the model; and S6, obtaining a land degradation degree classification result by using the trained land degradation degree classification model. The problems that in the prior art, only a single environment index is concerned, other multi-source environment factors are neglected, and a complex nonlinear relation existing in the actual process is neglected, so that result deviation is large, and great subjectivity exists are solved.

Description

Technical Field

[0001] The invention belongs to the field of resource environment and land management, and in particular relates to a land degradation degree classification method based on convolutional neural network and multi-source data. Background Art

[0002] Land degradation is a global environmental problem that poses a serious threat to ecosystem services, food security and human well-being. The United Nations Sustainable Development Goals (SDGs) list "zero land degradation by 2030" as one of the important goals. However, with population growth, urbanization, climate warming and irrational use of land resources, the scale and rate of land degradation are increasing, posing severe challenges to agricultural production, ecological security and sustainable economic and social development. Objectively classifying the degree of land degradation is the basis for formulating effective land management policies and restoration strategies, so it is extremely necessary to formulate accurate standards for the classification of land degradation.

[0003] The classification method of land degradation has gone through a development process from qualitative to quantitative, from single indicator to comprehensive multi-indicator. Early land degradation classification mainly relied on expert experience and field surveys. Although this method can obtain accurate point information, it is time-consuming and labor-intensive and easily affected by human accessibility, terrain, weather, and time cost. Therefore, it is impossible to quickly obtain distributed information on land degradation in a large area. With the development of earth observation technology, remote sensing technology has begun to be used for land degradation monitoring and classification. Based on this, the commonly used land degradation classification methods developed include single index method, empirical model method, comprehensive index method, and land use change method. The single index method uses single environmental indexes such as normalized difference vegetation index (NDVI), vegetation coverage, surface temperature, soil organic matter content and other indicators to reflect the degree of land degradation. The empirical model method uses empirical knowledge such as soil and water loss equation (RUSLE) and ecological environment quality index (RSEI) to measure the degree of land degradation using soil and water loss and ecological quality. The comprehensive index method uses expert scoring method (AHP method, analytic hierarchy process) or objective weighting (entropy weight) and classifies the degree of land degradation through weighted synthesis of multiple environmental indicators (such as vegetation index, soil moisture, rainfall, etc.). The land use change method studies land degradation based on land use / land cover change (LUCC) and determines the degree of land degradation by analyzing changes in land types such as agricultural land, forest land, grassland, and wasteland.

[0004] The existing land degradation classification methods provided an important theoretical and practical basis for solving the land degradation problem in early studies. Although the above classification methods have promoted the progress of land degradation degree classification methods, there are still some defects that cannot be ignored: 1) Only focusing on a single environmental index, ignoring the comprehensive influence of other multi-source environmental factors such as soil characteristics, topography, climate and other factors; 2) It is usually assumed that land degradation influencing factors such as rainfall, soil and other parameters change linearly, ignoring the complex nonlinear relationship in the actual process, resulting in large deviations in the results; 3) The classification standards and classification process of land degradation degree are still extremely subjective, and no unified, cross-regionally comparable degradation degree classification characterization standards and technical processes have been formed.

[0005] To solve the above problems, it is urgent to develop a new objective classification method for land degradation degree that is objective and universal, integrates multi-source data, and uses nonlinear modeling to cope with the increasingly complex land degradation problem and improve classification accuracy and efficiency. Summary of the invention

[0006] The purpose of the embodiment of the present invention is to provide a land degradation degree classification method based on convolutional neural network and multi-source data, so as to solve the problem that the prior art only focuses on a single environmental index, ignores other multi-source environmental factors, and ignores the complex nonlinear relationship in the actual process, resulting in large deviations in the results and great subjectivity.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0008] The land degradation degree classification method based on convolutional neural network and multi-source data includes the following steps:

[0009] S1: Determine the criteria for the degree of land degradation;

[0010] S2: Obtain land degradation degree sample data and classify the sample points according to the land degradation degree standard;

[0011] S3: Obtain environmental variable data and perform preprocessing;

[0012] S4: uniformly process the classified land degradation degree sample point data and the pre-processed environmental variable data to obtain sample data;

[0013] S5: Construct a land degradation degree classification model and train the model;

[0014] S6: Use the trained land degradation degree classification model to obtain the land degradation degree classification results.

[0015] Furthermore, the S1 includes:

[0016] The soil profile hierarchy is used to classify the land degradation degree standards into:

[0017] No degradation: The soil profile structure is a complete ABCR layer structure.

[0018] Mild degradation: In the soil profile structure, layer A is completely lost and layer B is exposed.

[0019] Moderate degradation: In the soil profile structure, the A and B layers are completely lost, and the C layer is exposed.

[0020] Strength degradation: In the soil profile hierarchy, the A, B, and C layers are lost and the R layer is exposed.

[0021] Further, the S2 includes:

[0022] S21: Collect soil stripping degree data of n land degradation degree sampling points and record the spatial coordinates of the sampling points;

[0023] S22: Based on the soil stripping degree data of the sample points and combined with the land degradation degree standards, the land degradation degree of each sample point is classified.

[0024] Further, the S3 includes:

[0025] S31: Get Sentinel-2 data and preprocess it:

[0026] Acquire the atmospheric top apparent reflectance TOA data of Sentinel-2; perform median synthesis processing on the Sentinel-2 remote sensing data set; and then calculate the spectral index, which includes soil adjusted vegetation index SAVI, normalized difference vegetation index NDVI, enhanced vegetation index EVI, total greenness index SGI, normalized burn ratio NBR, leaf area index LAI, clay mineral index CMR, iron ore ratio index, iron oxide ratio index, and regional burn index;

[0027] S32: Get Sentinel-1 data and preprocess it:

[0028] Acquire Sentinel-1 ground range detection GRD data and perform median synthesis; the ground range detection GRD data contains the backscatter coefficient in the polarization mode of vertical transmission and vertical reception, namely the VV backscatter coefficient, and the backscatter coefficient in the polarization mode of vertical transmission and horizontal reception, namely the VH backscatter coefficient; then perform differential normalization on the VV backscatter coefficient and the VH backscatter coefficient to obtain the radar vegetation index RVI;

[0029] S33: Get digital elevation model DEM data and preprocess it:

[0030] Obtain DEM data, perform terrain analysis on the DEM data, and obtain parameters such as altitude, slope, slope aspect, terrain curvature, terrain slope index, surface undulation, runoff dynamic index, surface roughness, terrain moisture index, and slope length.

[0031] Furthermore, the S3 also includes:

[0032] S34: Obtain soil data, climate data, and land use type data and perform preprocessing:

[0033] Obtain soil data, including soil organic matter content, soil sand content, soil silt content, and soil clay content; climate data includes the annual average temperature and annual average precipitation from the current year to a year ago; obtain land use type data;

[0034] The data acquired in S31 to S34 are all environmental variable data, and the forms of these environmental variable data are all in raster form;

[0035] S35: Data rasterization;

[0036] The soil parent material type data and geological type data were obtained and rasterized;

[0037] S36: Data standardization;

[0038] The raster data of all environmental variables were resampled to the same resolution; the land use type data, geological type data, and soil parent material type raster data were converted into numerical form, and the raster data of other environmental variables except these three environmental variables were normalized.

[0039] Further, the S4 includes:

[0040] After unifying the rasterized environmental variable data and the sample point land degradation degree classification data obtained by S2 into the projection coordinate system, all the environmental variable data corresponding to each sample point position are extracted to obtain sample data with a dimension of m×n, where m is the number of variables, i.e., the sum of the number of environmental variables and the number of land degradation degree variables, and n is the number of sample points.

[0041] Further, the S5 includes:

[0042] S51: Encode the land degradation degree variable into a vector form and preset the model learning rate, batch, and number of iterations:

[0043] No degradation is [1, 0, 0, 0], mild degradation is [0, 1, 0, 0], moderate degradation is [0, 0, 1, 0], and severe degradation is [0, 0, 0, 1];

[0044] S52: Build and train the model;

[0045] The land degradation degree classification model consists of input layer, output layer, convolution layer, pooling layer, fully connected layer, flattening layer, and Dropout layer;

[0046] In the input layer, the input image size is 64x64, and each pixel has 32 channels;

[0047] In each convolution layer, the size of the convolution kernel is 3x3, and the number of convolution kernels is 64, 128, or 256;

[0048] In each pooling layer, 2x2 maximum pooling is used;

[0049] In the output layer, the Softmax activation function is used to output the classification results of land degradation degree;

[0050] The sample data obtained in S4 is divided into a training set and a validation set, and the model is trained using the training set, the cross entropy loss function, and the Adam optimizer.

[0051] The beneficial effects of the present invention are:

[0052] 1. Innovation and objectivity of the definition of land degradation degree. The present invention defines the degree of land degradation as the relative stage of land degradation development, and uses the soil profile hierarchy (A layer, B layer, C layer and R layer) exposed after the in-situ soil degradation as an intuitive representation. This method has the following advantages: 1) Objectivity: Based on the observable soil profile structure, it reduces subjective judgment and has a clear physical meaning; 2) Comparability: It provides a unified standard for the classification of land degradation degrees in different regions and periods; 3) Clear physical meaning: It intuitively reflects the actual degree of soil degradation, which is easy to understand and apply; 4) It helps to establish a global unified land degradation classification system and improve the comparability and reliability of classification results.

[0053] 2. Comprehensiveness and systematicness of multi-source data fusion. The present invention comprehensively utilizes multi-source spectral optical remote sensing, synthetic aperture radar, DEM data obtained by satellite platforms, and climate, soil, and geological information obtained by ground monitoring systems as factors affecting the degree of land degradation. This multi-source data fusion method has the following advantages: 1) Comprehensiveness: It covers a variety of factors that affect land degradation. 2) Complementarity: Different data sources complement each other and provide more comprehensive surface information. 3) Multi-scale: It combines the large-scale observations of satellite remote sensing and the fine observations of ground monitoring to improve the accuracy and reliability of classification.

[0054] 3. Innovation and efficiency of deep learning algorithms. The present invention adopts an innovative convolutional neural network architecture, which has the following advantages: 1) Automatic feature extraction: Deep learning can automatically learn complex feature representations and reduce manual feature engineering. 2) Nonlinear modeling capabilities: It can capture complex nonlinear relationships in the land degradation process. 3) End-to-end learning: Directly learn from raw data to the final degree of degradation, reducing cumbersome intermediate steps. The accuracy and efficiency of land degradation degree classification are significantly improved, especially in large-scale applications. 4) High precision: Deep learning algorithms can effectively capture complex nonlinear relationships and improve classification accuracy.

[0055] 4. The method of the present invention designs an automated classification process, which is suitable for land degradation classification in large-scale areas. This feature has the following advantages: 1) High efficiency: Automated processing of large amounts of data significantly improves classification efficiency. 2) Consistency: Unified classification standards and methods ensure consistency of results. 3) Extensibility: The method is universal and can be extended to different regions. It provides a feasible technical solution for land degradation monitoring and classification at regional or even global scales.

[0056] In summary, the method of the present invention uses multi-source data fusion and deep learning algorithms to make innovations and breakthroughs in the classification of land degradation degrees, significantly improving the scientificity, accuracy and practicality of land degradation classification. These advantages and positive effects make this method of important application value in achieving the goal of zero growth of land degradation and promoting regional sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 It is a flow chart of the present invention.

[0059] Figure 2 This is a schematic diagram of land degradation degree standards.

[0060] Figure 3 This is the sample distribution map of land degradation degree survey.

[0061] Figure 4 This is a ranking diagram of the importance of land degradation environmental factors in the case area of ​​the present invention.

[0062] Figure 5 This is the distribution map of the classification results of land degradation degree in the case area of ​​the present invention. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] The complete steps of the method of the present invention are as follows:

[0065] See also Figure 1 , the steps of the embodiment of the present invention are as follows:

[0066] S1: Determine the criteria for the degree of land degradation.

[0067] See also Figure 2 This paper defines the degree of land degradation as the relative stage of land degradation development for the first time, and uses the current soil profile hierarchy (A layer, B layer, C layer and R layer) exposed after the in-situ soil body is degraded as an intuitive representation. It is specifically divided into four degradation degrees:

[0068] No degradation: Complete ABCR layer structure,

[0069] Mild degradation: Layer A is completely lost and layer B begins to be exposed.

[0070] Moderate degradation: B layer is completely lost, and C layer is exposed over a large area.

[0071] Strength degradation: Most of the C layer is lost and the R layer begins to be exposed.

[0072] The detailed description of each degradation level is shown in Table 1.

[0073] Table 1 Land degradation degree and characteristics

[0074]

[0075] This method can be used to obtain the degree of land degradation through field surveys. This definition method is based on observable soil profile structure, which reduces subjective judgment; it provides a unified standard for the classification of land degradation degrees in different regions and periods; it intuitively reflects the actual degree of soil degradation, which is easy to understand and apply; it helps to establish a global unified land degradation classification system and improve the comparability and reliability of classification results.

[0076] S2: Acquisition of sample data of land degradation degree and sample classification. Through field survey, obtain sample data of land degradation degree survey in the case area. By observing the degree of soil detachment and combining the land degradation degree classification standard proposed in S1, classify the land degradation degree of each sample point, and use the determined land degradation degree as the dependent variable in the land degradation degree classification model. The specific steps are described as follows:

[0077] S21: See Figure 3 Through field investigations, 297 land degradation survey sample data from the case area in 2020, namely the soil detachment degree information of the sample points, were collected, and the spatial coordinate information of the sample points was recorded using a handheld GPS instrument.

[0078] S22: Based on the soil stripping degree information of the sample points in the field survey and the proposed land degradation degree classification standard, the land degradation degree of each sample point is classified, and the land degradation degree of the sample is used as the model dependent variable.

[0079] S3: Factors affecting the degree of land degradation, namely, acquisition and processing of environmental data.

[0080] Considering that land degradation is caused by the combined effects of natural and human activities, multispectral remote sensing data (Sentinel-2 satellite), synthetic aperture radar (Sentinel-1 satellite), DEM, soil properties, climate, land use, and geological type data of the case area were obtained as independent variables of the land degradation degree classification model; the specific steps are described as follows:

[0081] After preprocessing these environmental data, the independent variables of the model are shown in Table 2.

[0082] Table 2 Independent variables used in the land degradation degree classification model

[0083]

[0084]

[0085] S31: Acquisition and processing of Sentinel-2 data.

[0086] All Sentinel-2 atmospheric top apparent reflectance TOA product data with spatial resolutions of 10, 20, and 60 m in the case area in 2020 were obtained on the Google Earth Engine remote sensing cloud platform. The data had been processed by radiometric calibration, atmospheric correction, and geometric correction. Cloud cover statistics were performed on the images, and only data with cloud cover less than 5% were retained. Median synthesis processing was performed on all eligible Sentinel-2 remote sensing datasets to reduce cloud cover and data noise. Then, 10 spectral indices were extracted through band calculation, including soil adjusted vegetation index (SAVI), normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), sum green index (SGI), normalized burn ratio (NBR), leaf area index (LAI), clay minerals ratio (CMR), and iron ore ratio index (Ferrous mineral ratio). Minerals Ratio (FMR), Iron Oxide Ratio (IOR), Burn area index (BAI). These spectral indices can invert land cover, vegetation growth, soil characteristics, and roughness.

[0087] S32: Acquisition and processing of Sentinel-1 data.

[0088] All Sentinel-1 GRD product data of the case area in 2020 after speckle filtering, terrain correction, etc. were obtained on the Google Earth Engine remote sensing cloud platform; it includes the backscatter coefficients in VV and VH polarization modes; all Sentinel-1 synthetic aperture radar data are median synthesized to reduce noise and obtain clearer images, and then the backscatter coefficients in VV and VH polarization modes are differentially normalized to calculate the radar vegetation index (RVI).

[0089] S33: Processing DEM data.

[0090] The STRM DEM v3.0 product DEM data with a spatial resolution of 30m in the case area was downloaded through the Google Earth Engine remote sensing cloud platform, and then the regional DEM data was used for terrain analysis using SAGA GIS software to extract terrain parameters such as altitude, slope, slope aspect, terrain curvature, terrain slope index, surface undulation, runoff dynamic index, surface roughness, terrain moisture index, and slope length.

[0091] S34: Acquisition and processing of soil, climate data, and land use types.

[0092] Soil attribute raster data with a spatial resolution of 90 m was obtained from the Resource and Environmental Science Data Center of the Chinese Academy of Sciences, including soil organic matter, sand, silt and clay content data; meteorological data is 1 km × 1 km raster data of annual average temperature and annual average precipitation for many years, which needs to be able to reflect the climate background characteristics. Generally, the longer the time series, the better. In this embodiment, 1 km × 1 km raster data of annual average temperature and annual average precipitation from this time to 43 years ago is selected;

[0093] Obtain land use type raster data at 30m spatial resolution;

[0094] The above environmental variable data are all in the form of raster data.

[0095] S35: Data rasterization.

[0096] Obtain soil parent material type data; obtain geological type data of the region from the China Geological Survey Center, with a scale of 1:200,000.

[0097] The obtained soil parent material type and geological type vector geological type data were rasterized through the ArcGIS 10.8 platform.

[0098] S36: Data standardization.

[0099] In ArcGIS 10.2 software, the raster data of all 32 environmental variables were resampled to a spatial resolution of 30 m using the cubic method. In addition, the raster data of land use, geological type, and soil parent material type were converted into numerical form through dummy variable processing, and other environmental variables were normalized.

[0100] S4: Construct sample data.

[0101] Using ArcGIS 10.8 software, the above 32 environmental variable data (raster spatial data) and the classified land degradation degree sample data (spatial points are data) were unified into the UTM-WGS-1984 projection coordinate system, and then all the environmental variable information corresponding to each sample point position was extracted, thereby obtaining sample data with a dimension of 33*297, where 33 is the number of variables and 297 is the number of samples.

[0102] S5: Model construction and accuracy classification.

[0103] The 297 samples of land degradation degree in the case area are used as dependent variables, and the standardized spectral remote sensing data (Sentinel-2 satellite), synthetic aperture radar (Sentinel-1 satellite), DEM, soil properties, climate, land use, and geological type data in S36 are used as independent variables; a land degradation degree classification model based on convolutional neural network is first constructed on the Python platform; and the accuracy of the classification model and classification results is quantitatively evaluated. The specific steps include the following:

[0104] S51: Digitally express the degree of land degradation and preset the model learning rate, batch, and number of iterations.

[0105] The degree of land use degradation is encoded as a one-hot vector, where no degradation is [1, 0, 0, 0], mild degradation is [0, 1, 0, 0], moderate degradation is [0, 0, 1, 0], and severe degradation is [0, 0, 0, 1]; this facilitates the operation of the convolutional neural network algorithm.

[0106] The cross entropy loss function and Adam optimizer are used in the model training process. The initial learning rate is 0.001, the batch size is 32, the number of iterations is 200, the learning rate scheduling strategy is to halve the learning rate every 50 epochs, and the early stopping strategy is to stop training if the loss does not improve within 20 epochs.

[0107] S52: Build the model and train the model.

[0108] The obtained sample data with a dimension of 33*297 were randomly divided into a training set (209) and a validation set (88) in a ratio of 7:3, which were used for model training and model accuracy verification respectively.

[0109] Based on the data structure characteristics of this embodiment, a convolutional neural network classification model is constructed, and the structure is shown in Table 3.

[0110] Table 3 Convolutional neural network classification model parameters of this embodiment

[0111]

[0112]

[0113] The model is trained using the training set. The land degradation degree classification model based on convolutional neural network makes full use of the advantages of deep learning. It can automatically extract features from multi-source data, establish complex nonlinear relationships, and achieve accurate and objective land degradation degree classification.

[0114] S6: Obtain the predictive classification results of land degradation degree.

[0115] The land degradation degree classification model based on convolutional neural network that has been trained in S5 is applied to the 32 environmental variable raster datasets with a spatial resolution of 30m in the case area and classified to obtain the predictive classification results of land degradation degree in the case area in 2020, such as Figure 5 .

[0116] Model accuracy:

[0117] The validation set is used to validate the model trained in S52. The overall accuracy (OA), Kappa coefficient (k), producer accuracy (PA), and user accuracy (UA) are used to quantitatively evaluate the reliability of the model. OA is a relatively comprehensive indicator, which is expressed as the total number of correctly classified samples divided by the total number of validation samples:

[0118]

[0119] Where N is the total number of validation samples, i refers to each category in the confusion matrix. In the confusion matrix, each category i corresponds to a row and a column of the matrix, r is the total number of categories, and x ii is the number of samples with the same true category as the classified category, x +i represents the number of samples classified into this category, that is, the sum of the i-th row; x i+ Represents the number of samples that actually belong to this category, that is, the sum of the i-th column.

[0120] As shown in Table 4, the accuracy of the land degradation degree classification model integrating multi-source data and convolutional neural network is OA 82.02% and Kappa is 73.30%.

[0121] Table 4 Confusion matrix of land degradation degree (N=89)

[0122]

[0123]

[0124] Figure 5The classification results provide a comprehensive analysis of the land degradation status of the case area. The distribution of land degradation degree: mild degradation accounts for 45.03%, moderate degradation accounts for 36.31%, severe degradation accounts for 9.54%, and extremely severe degradation accounts for 9.12%. Spatial distribution characteristics: The central mountainous area, namely the valley area, is mainly non-degraded and mildly degraded; moderate and severe degradation occurs more frequently on gentle slopes and gentle hills, and extremely severe degradation is mainly distributed on ridges and slope tops where vegetation growth is poor and the surface is gently sloping.

[0125] Based on the land degradation degree classification model based on convolutional neural network trained in S5, the permutation importance method is used to calculate the importance of a single environmental variable to the model, which can be expressed as the importance ranking of the factors affecting land degradation in the case area.

[0126] The importance scores of all environmental variables were normalized to the range of 0-1 for comparison. Figure 4 .

[0127] Figure 4 The normalized difference vegetation index, soil organic matter content, multi-year average precipitation, leaf area index, and radar vegetation index show high importance, with the importance being 15.06%, 13.90%, 12.10%, 11.88%, and 10.59% respectively; reflecting the important indicative role of vegetation coverage and soil quality on land degradation. This analysis result provides a comprehensive perspective to understand the relative importance of different environmental factors in the process of land degradation. It can not only guide us to use these variables more targetedly in the model, but also provide a scientific basis for land management and ecological restoration.

[0128] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A land degradation degree classification method based on convolutional neural network and multi-source data, characterized in that: The following steps are involved: S1: Determine the criteria for the degree of land degradation; S2: Obtain land degradation degree sample data and classify the sample points according to the land degradation degree standard; S3: Obtain environmental variable data and perform preprocessing; S4: uniformly process the classified land degradation degree sample point data and the pre-processed environmental variable data to obtain sample data; S5: Construct a land degradation degree classification model and train the model; S6: Use the trained land degradation degree classification model to obtain the land degradation degree classification results.

2. The land degradation degree classification method based on convolutional neural network and multi-source data according to claim 1 is characterized in that: The S1 includes: The soil profile hierarchy is used to classify the land degradation degree standards into: No degradation: The soil profile structure is a complete ABCR layer structure. Mild degradation: In the soil profile structure, layer A is completely lost and layer B is exposed. Moderate degradation: In the soil profile structure, the A and B layers are completely lost, and the C layer is exposed. Strength degradation: In the soil profile hierarchy, the A, B, and C layers are lost and the R layer is exposed.

3. The land degradation degree classification method based on convolutional neural network and multi-source data according to claim 2 is characterized in that: The S2 includes: S21: Collect soil stripping degree data of n land degradation degree sampling points and record the spatial coordinates of the sampling points; S22: Based on the soil stripping degree data of the sample points and combined with the land degradation degree standards, the land degradation degree of each sample point is classified.

4. The land degradation degree classification method based on convolutional neural network and multi-source data according to claim 1 is characterized in that: The S3 includes: S31: Get Sentinel-2 data and preprocess it: Acquire the atmospheric top apparent reflectance TOA data of Sentinel-2; perform median synthesis processing on the Sentinel-2 remote sensing data set; and then calculate the spectral index, which includes soil adjusted vegetation index SAVI, normalized difference vegetation index NDVI, enhanced vegetation index EVI, total greenness index SGI, normalized burn ratio NBR, leaf area index LAI, clay mineral index CMR, iron ore ratio index, iron oxide ratio index, and regional burn index; S32: Get Sentinel-1 data and preprocess it: Acquire Sentinel-1 ground range detection GRD data and perform median synthesis; the ground range detection GRD data contains the backscatter coefficient in the polarization mode of vertical transmission and vertical reception, namely the VV backscatter coefficient, and the backscatter coefficient in the polarization mode of vertical transmission and horizontal reception, namely the VH backscatter coefficient; then perform differential normalization on the VV backscatter coefficient and the VH backscatter coefficient to obtain the radar vegetation index RVI; S33: Get digital elevation model DEM data and preprocess it: Obtain DEM data, perform terrain analysis on the DEM data, and obtain parameters such as altitude, slope, slope aspect, terrain curvature, terrain slope index, surface undulation, runoff dynamic index, surface roughness, terrain moisture index, and slope length.

5. The land degradation degree classification method based on convolutional neural network and multi-source data according to claim 4 is characterized in that: The S3 further includes: S34: Obtain soil data, climate data, and land use type data and perform preprocessing: Obtain soil data, including soil organic matter content, soil sand content, soil silt content, and soil clay content; climate data includes the annual average temperature and annual average precipitation from the current year to a year ago; obtain land use type data; The data acquired in S31 to S34 are all environmental variable data, and the forms of these environmental variable data are all in raster form; S35: Data rasterization; The soil parent material type data and geological type data were obtained and rasterized; S36: Data standardization; The raster data of all environmental variables were resampled to the same resolution; the land use type data, geological type data, and soil parent material type raster data were converted into numerical form, and the raster data of other environmental variables except these three environmental variables were normalized.

6. The land degradation degree classification method based on convolutional neural network and multi-source data according to claim 3 or 5, characterized in that: The S4 includes: After unifying the rasterized environmental variable data and the sample point land degradation degree classification data obtained by S2 into the projection coordinate system, all the environmental variable data corresponding to each sample point position are extracted to obtain sample data with a dimension of m×n, where m is the number of variables, i.e., the sum of the number of environmental variables and the number of land degradation degree variables, and n is the number of sample points.

7. The land degradation degree classification method based on convolutional neural network and multi-source data according to claim 6 is characterized in that: The S5 includes: S51: Encode the land degradation degree variable into a vector form and preset the model learning rate, batch, and number of iterations: No degradation is [1, 0, 0, 0], mild degradation is [0, 1, 0, 0], moderate degradation is [0, 0, 1, 0], and severe degradation is [0, 0, 0, 1]; S52: Build and train the model; The land degradation degree classification model consists of input layer, output layer, convolution layer, pooling layer, fully connected layer, flattening layer, and Dropout layer; In the input layer, the input image size is 64x64, and each pixel has 32 channels; In each convolution layer, the size of the convolution kernel is 3x3, and the number of convolution kernels is 64, 128, or 256; In each pooling layer, 2x2 maximum pooling is used; In the output layer, the Softmax activation function is used to output the classification results of land degradation degree; The sample data obtained in S4 is divided into a training set and a validation set, and the model is trained using the training set, the cross entropy loss function, and the Adam optimizer.

Images